"What if you were perfectly prepared for the next six seconds, what would you do?”

Back in 1992, two exciting things happened in my life. First of all, my second child and eldest daughter was born. Then, I received my NASA manned space flight launch room certification as a certified Ground Launch Sequencer (GLS) programmer with the responsibility of programing and overseeing the GLS math model that runs the shuttle countdown from T-9 through liftoff.  I realize the birth of my daughter is much easier to grasp than all of that space gibberish, but to my 28-year-old self, they were both crowning achievements. The way I was supporting my little family was by ensuring that the lines of code I edited, created, and complied, resulted in a safe and successful launch of the Space Shuttle.

The GLS is an amazing piece of software that originated in the late 1970’s and was first used in 1981 to launch STS-1, Columbia with Bob Young and Bob Crippen on board for the first flight of the space shuttle program. The next 134 launches depended on the GLS interrogating over 75,000 functional designators some sampling at 1,000 samples/sec to decide if it was safe to move the launch countdown forward. Each mission changed the flight dynamics of the vehicle, so each mission required updates and changes to the GLS tapestry of code, so it’s generic set of “business rules” could be modified successfully to accommodate the particular nuances of the next STS mission profile. Understanding that at exactly at T-2m, 55sec it’s time to pressurize the 146,181 gallons of liquid oxygen, not before then, not after then is the very nature of the GLS design. If you do this too soon, or too late¬– well, let’s just say disastrous things can happen.   The GLS system was one part code, but the other part is Launch Control Room Certified GLS operators, whose job it is to study and then improve the GLS code for the next launch. This is “Human Learning” in action. Taking months between launches to prepare new code. So how do humans teach a machine to learn, when they learn within nanoseconds? First, we start with Brand, focus on the journey, use the right data, and then deliver the perfect interaction every six seconds.

Of course in this blog the question looms, what does being a GLS programmer have to do with Customer Experience and the Machine Learning AI bots being deployed to control your companies every CX touch-point?  Well, I submit that these four steps connect everything:

1)    BRAND - Focus on your Brand

  • Why: As GLS programmers our brand promise was to deliver astronauts and cargo into low earth orbit safely.

2)    JOURNEY - Understand the Journey

  • When: We navigated a sequence of events that led to a successful outcome, and constantly monitored for known and unknown pain points.

3)    DATA - Measure what you manage, Manage what you measure

  • What: By measuring all forms of structured, unstructured, activities, and outcomes we could leverage the Launch Processing System multiple data sets and created a data lake.

4)    INTERACTION - Every interaction matters

  • How: Each interaction is an opportunity to learn and improve the next interaction. The Ground Launch Sequencer engine learned and improved the safety and predictability of each subsequent launch by closely facilitating the next ‘right’ step.

Each of these steps ties closely to creating a successful machine learning AI bot that can be trusted to operate your companies CX touch points successfully.               


The first step deconstructs your Brand into distinct, and tangible elements. You have to ask yourself what exactly is your company’s promise, and how can you quantify those qualities? Boiling everything down into a math equation, so that the machine learning/AI bot can create a business rule that will instruct its next action.

Let’s consider Southwest Airlines brand promise: “We exist to connect people to what’s important in their lives.” Their brand promise centers on five key characteristics: “friendly,” “reliable,” “low-cost,” “above & beyond,” and “hungry & humble.” These are all great ideas, but how does a company measure “friendly” or “hungry & humble?”  Over the past ten years, it has become easier than ever to measure things– causing many companies to ‘drown in data.'   The key to measuring exactly what you want, like being friendly, is to listen for the right signals at the right time. Start by drawing a 2x2 box and brainstorming Structured vs. Unstructured data, along with Activities/Events vs. Outcomes.

So, focusing on the brand promise of “Friendly” and how to measure it:

Screen Shot 2017-08-25 at 7.17.51 PM.png

By breaking every aspect down, we can capture the data needed to establish the mathematical baselines and norms that can be used to inform our machine learning AI bots.


In the second step focuses the customer journey and asks the question “What does a successful customer look like and how will you recognize them” needs to be answered. Understanding what a successful customer life cycle looks like will help you tailor your AI when creating successful interactions. Just like plotting a course cross-country, think about the waypoints that customers need to pass through to tell if they are on track to success. Waypoints can be anything from hitting “like” on Facebook to completing a shopping cart, and the hundreds of other ways customers grow close to the brands they love. To discover the key waypoints, we need to focus on the six individual sections of the life cycle and determine the key data points needed to measure waypoint navigational success.


Now that we understand the Why and the When we need to focus now on the What. In this step, we work to connect the right data ‘puddles’ a company creates in their marketing, sales, service, product development departments as well as other data streams outside of the company across different market sectors that can be brought together to create one massive data lake. Mapping the potential data sources from the key performance indicators captured in Steps 1 and two will connect all the required data tributaries to your data lake to teach and inform your machine learning. Last blog I talked about how Machine Learning and Artificial Intelligence can only learn based on what data is given to them. When pulling from small data puddles built by different sections of a company, AI can get an incomplete picture of what your brand promise is, or how a successful customer looks. Finding the connections between the relevant data puddles and giving the machines a larger pool to learn from is going to result in more successful bots. This will provide the foundation needed to focus on the “next 6 seconds” problem.      


The last step asks, “In the next 6 seconds - what is the next “Best” step for our customer to take, so that they feel our brand promises, as they successfully complete their interaction, while moving towards the next customer lifecycle waypoint?” By pulling from all the activities and outcomes captured in steps 1 and 2, an AI bot learns from the data lake and can then decide what the best next step for a customer, even if it means handing over control to a customer experience representative.

When launching a space shuttle the code has to understand the micro-interactions of the main engine start at T-6.8 seconds, and verify successful ignition & thrust before T-0. Understanding that the last 6 seconds of launch are built on top of a launch countdown that started over 42 hours ago (as well as the previous 100+ missions) is key to appreciating how important interaction design and monitoring are when crafting a world class AI bot.  There may be a high impact issue that only a human can resolve for the customer and bots should be able to make that transition smoothly based on machine learning validated business rules.

Just like with parenting, teaching an AI enabled customer experience compels us to start with our heart, being mindful of the stages through childhood, observing the signs of progress and being present as much as possible all the key events that make up the life of a young adult. If your lucky, like my younger self back in 1992, you will find yourself 25 years later with a daughter who is finishing her masters degree in writing, while planning her wedding for next year.   Brand/Heart, Journey/Mind, Data/Awareness, and Interaction/Moments– these are the right four steps to first focus on when starting down the path of automating your customer experience.



Image credit: Jetsons, 1962 Hanna-Barbera